Hi Dave, Laurent, On Tue, Aug 22, 2023 at 06:58:47PM +0100, Dave Stevenson wrote: > Hi Laurent > > On Mon, 21 Aug 2023 at 23:30, Laurent Pinchart > <laurent.pinchart@xxxxxxxxxxxxxxxx> wrote: > > > > There is no need to stop streaming at system suspend time, or restart it > > when the system is resumed, as the host-side driver is responsible for > > stopping and restarting the pipeline in a controlled way by calling the > > .s_stream() operation. Drop the system suspend and resume handlers, and > > simplify the .s_stream() handler as a result. > > I'll believe you, but the docs for power management in camera sensor > drivers [1] state > "Please see examples in e.g. drivers/media/i2c/ov8856.c and > drivers/media/i2c/ccs/ccs-core.c. The two drivers work in both ACPI > and DT based systems." > > Looking at CCS we find the suspend hook stopping streaming [2], and > resume hook starting it [3]. Same in ov8856 [4]. > > Could you reference the documentation that states that the host-side > driver is responsible for starting and stopping? Is this an ACPI vs DT > difference? There's no difference between DT and ACPI, no. Starting streaming on resume from system suspend is haphazard as there's no guarantee on the timing of resuming devices (if the suspend and resume independently), or similarly if the receiver driver explicitly starts streaming, there's no guarantee the sub-device driver has resumed. So I think we'd need more than what currently exists on the framework side to do this reliably --- at least when it comes to CSI-2. -- Regards, Sakari Ailus
